The human complement inhibitor Sushi Domain-Containing Protein 4 (SUSD4) expression in tumor cells and infiltrating T cells is associated with better prognosis of breast cancer patients

Tissue samples obtained from a cohort of 144 women diagnosed with breast cancer in Skåne, Sweden [6] were stained with rabbit anti-SUSD4 (home-made) as previously validated and described [1]. Ethical permission was obtained from the Lund University Regional Ethics Board, ref. no. 445/2007 whereby written consent was not required and patients were offered the option to opt out. The intensity of the SUSD4-specific signal in tumor cells was scored 0 (no expression), 1 (low expression) or 2 (high expression) independently by two scientists and one experienced clinical pathologist, who were all blinded with regard to clinical information. For statistical analyses, the scores were grouped into SUSD4 negative (score 0) and SUSD4 positive tumors (scores 1–2). In the stroma of the tumors, SUSD4 positive tumor-infiltrating cells were detected. The cells were counted for the whole tissue section and grouped into 0–15 (low) or 16–100 (high) SUSD4⁺ cells/section. Kaplan-Meier analyses and Breslow tests were used to determine the effect of SUSD4 expression by tumor cells or infiltrating cells on cancer-specific survival and recurrence-free survival. Uni- and multivariable Cox proportional hazard models based on SUSD4 expression were used to determine hazard ratios (HR) for cancer-specific death. Immunohistochemical data regarding hormone receptor status, Ki-67 and human epidermal growth factor receptor 2 (HER2) expression were available from previous studies [6, 7]. Definitions of estrogen receptor (ER) and progesterone receptor (PR) negativity followed current Swedish clinical guidelines (<10 % positive nuclei). Ki-67 status was assessed based on the percentage of positively stained nuclei and dichotomized into ≤25 % and >25 %. HER2 status was assessed by semiquantitative analysis according to a standard protocol [8]. Specimens were grouped as weakly (scores 0–2) and strongly expressing HER2 (score 3). Any differences in the distribution of clinical parameters and SUSD4 expression were calculated for tumor cells and tumor-infiltrating cells by using 2-tailed Mann–Whitney U tests. Exact p-values <0.05 were considered statistically significant. The calculations were performed using SPSS Statistics v. 22 (IBM). The original slides were scanned with an Aperio ScanScope slide scanner and representative pictures (40X magnification) were obtained in the ImageScope software (Aperio).

Fresh frozen mammary tissues (normal, n = 32 and tumour, n = 127) from patients with breast cancer were collected immediately after surgery, under the research ethics approval from the Southeast Wales Research Ethics Committee and with informed consent, and stored at −80 °C until used. Patients were followed up routinely in the clinics with a median follow-up at 120 months, and their clinical characterisation was published previously [9]. The tissues were homogenized and total RNA was purified. cDNA was synthesized from 1 μg RNA according to the manufacturer’s specifications (AbGene). A qPCR was set up as described in [9], using the following primers for SUSD4: 5’-AAAACCTTATCTGGTCGTC-3’ and 5’-ACTGAACCTGACCGTACATCTCCGTGACTCACCATT-3’. A standard of cytokeratin-19 (CK19) was run simultaneously using primers 5’-CAGGTCCGAGGTTACTGAC-3’ and 5’-ACTGAACCTGACCGTACACACTTTCTGCCAGTGTGTCTTC-3’. The standard was used to obtain the transcript levels. The data were analyzed by Kaplan-Meier followed by Breslows test to determine if SUSD4 transcript levels affected cancer-specific survival or recurrence free survival. SUSD4 transcript levels were correlated to clinical parameters using Mann–Whitney U tests.

Breast cancer cell lines MDA-MB-231 and BT20 (American Type Culture Collection, ATCC) were cultured in DMEM high glucose (Thermo Scientific) medium supplemented with 10 % fetal bovine serum (FBS), penicillin and streptomycin. Cells were frozen immediately after re-cultivation of the original aliquot, and all the experiments were performed on cultures originating from these secondary aliquots within no more than 5 passages. Cells were Mycoplasma negative and tested monthly for contamination with the VenorGEM Classic kit (Minerva Biolabs). Although SUSD4 is predicted to be expressed as two isoforms, we focused this study only on the cancer-related functions of the membrane-bound SUSD4a, which is the isoform easily detectable at protein level. Full-length SUSD4a [1] was cloned into the pcDNA3 vector (Life technologies) using restriction sites EcoRI and XhoI. The construct or empty vector (mock) were transfected to MDA-MB-231 and BT20 cells using lipofectamine 2000 (Life technologies) and clones were selected with G418 (Life technologies). Cell pellets were collected and RNA was purified with the RNeasy kit (Qiagen). cDNA was synthesized from 1 μg RNA by using 2.5 μM oligo(dT) primer, 24 U RnaseOUT, and 200 U Superscript III reverse transcriptase (Life technologies). A qPCR was set up using 10 ng cDNA/well in triplicate for each sample. Specific primers detecting SUSD4a (Hs01042141_m1), cyclophilin A (Hs99999904_m1), TATA-box binding protein (Hs00427621_m1), and hypoxanthine phosphoribosyltransferase 1 (HPRT-1; Hs99999909_m1) were bought from Applied Biosystems. SUSD4a expression relative to the geometrical mean of the three references was calculated according to the ΔCt method [10].

SUSD4a protein expression was analysed by flow cytometry and western blot. For flow cytometry, 200 000 cells/well were incubated with 5 μg/ml anti-SUSD4 diluted in binding buffer (10 mM HEPES, 140 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 2 mM CaCl₂, 0.02 % w/v NaN₃, pH 7.2) for 1 hour at RT. The cells were washed in binding buffer, incubated with a secondary antibody conjugated to fluorescein isothiocyanate (FITC) for 30 min at RT, then resuspended in binding buffer and analysed by flow cytometry (Partec CyFlow Space flow cytometer) and the FlowJo software. For the western blot, lysates were run on a 12 % SDS-PAGE under reducing conditions. The gel was blotted (Trans-Blot Turbo, Bio-Rad) to a PVDF membrane, stained with 0.1 μg/ml anti-SUSD4 followed by a secondary antibody conjugated to horseradish peroxidase (HRP) and developed with ECL (Millipore).

Cells (6000 cells/well) were plated out in duplicates in four identical 96-well plates (Nunc). The plates were incubated for 0.5 h, 24 h, 72 h, or 96 h, before cell fixation with 4 % formaldehyde and staining with 0.5 % w/v crystal violet. Excess dye was washed away with tap water and the plate was left to dry over night. The dye was extracted with 10 % acetic acid and the absorbance was read at 540 nm using a microplate reader (Cary50Bio, Varian). The data were normalized to the highest value of each repetition.

A layer of matrigel (5 μg/well, BD Biosciences) was coated in quadruplicates in a 96-well plate. After drying and rehydration of the matrigel, cells (MDA-MB-231; 3x10⁴ cells/well and BT20; 5x10⁴ cells/well) were allowed to bind for 45 min at 37 °C. Unbound cells were removed by washing with BSS (680 mM NaCl, 15 mM KCl, 7 mM KH₂PO₄, 3.5 mM Na₂HPO₄, pH 7.2). The cells were fixed with 4 % formaldehyde and stained with 0.5 % w/v crystal violet as described above. After the plate was dry, two random pictures were taken of each well (40X objective; EVOS FL inverted microscope) and the cells were counted (ImageJ).

Cells were grown to confluency in a 6-well plate (Nunc). Two scratches were made per well with a sharpened yellow pipette tip. Pictures were taken at three positions (exactly the same position every time using a 10X objective) along each scratch at 0 h, 3 h and 6 h. The wound area was measured in ImageJ and the average of the 6 positions was used. The data were normalized to the wound area at 0 h for each sample.

Cells were resuspended in DMEM high glucose medium supplemented with 1 % FBS and placed in plain inserts (migration, 8 microns, BD Biosciences) in duplicates or inserts coated with matrigel (invasion, 8 microns, BioCoat, Corning) in singlet. The inserts were placed (before the addition of cells) in wells containing DMEM high glucose supplemented with 10 % FBS. Cells (MDA-MB-231; 5x10⁴ cells/well and BT20; 10x10⁴ cells/well) were left to migrate/invade for 22 h (MDA-MB-231) or 44 h (BT20). Cells that had moved to the underside of the inserts were fixed with formaldehyde and stained with crystal violet. Four pictures were taken of each inserts using a 40X objective and the number of migrating/invading cells was quantified with ImageJ.

Carcinoma-associated fibroblasts (CAFs) or control originating from breast tissue [11] were cultured O/N in triplicates in a 96-well plate (5x10³ cells/well). SUSD4a- or mock-transfected BT20 cells (100 cells/well) were added to each well and were cultured for 10 days. The cells were fixed with 70 % ethanol and stained with 0.1 % w/v toluidine blue (Sigma-Aldrich). Excess dye was removed with BSS and the plate was left to dry O/N. A picture of each well was taken using a 4X objective and the average cluster size of the colonies was quantified with ImageJ. Additional representative pictures were taken using a 40X objective.

CAFs were grown to confluency in quadruplicates in a 96-well plate before the addition of SUSD4a- or mock-transfected MDA-MB-231 or BT20 cells (30x10³ cells/well). The plate was incubated for 45 min and unbound cells were removed with BSS. Adherent cells were fixed with formaldehyde and stained with toluidine blue. The plate was rinsed with tap water and allowed to dry O/N at RT. Two random pictures were taken of each well using a 40X objective and the number of adherent cancer cells was counted using ImageJ.

Breast cancer TMAs (as described above) were double-stained for SUSD4 (2 μg/ml) and CD3 (10 μg/ml; clone OKT3; eBioscience), CD4 (8 μg/ml; clone 4B12; Dako) or CD8 (3 μg/ml; clone C8/144B; Dako) as described previously [12]. Pictures were taken with a Zeiss LCM 510 confocal microscope (20X objective).

Peripheral blood was drawn from healthy volunteers according to the permit of the local ethical committee in Lund, Sweden. Peripheral blood mononuclear cells (PBMCs) were isolated by centrifugation over a density gradient (Lymphoprep, Stemcell technologies). CD4⁺ or CD8⁺ T cells were purified from the PBMCs by using specific kits (Miltenyi Biotec). Wells (48-well plate) were coated O/N at 4 °C with a mix of 2 μg/ml anti-CD3 and 2 μg/ml anti-CD28 (clone CD28.2; eBioscience) diluted in PBS. The isolated T cells were either immediately frozen at −80 °C (fresh) or stimulated for 24 h in coated wells in RPMI medium (Thermo Scientific) supplemented with 10 % FBS and 5 U/ml IL-2 (Immunotools). Fresh and stimulated T cells were used for SUSD4a expression analysis simultaneously by both qPCR and western blot. The experiment was performed three times using blood from different donors.

RNA was purified (RNeasy, Qiagen) and cDNA was synthesized as described above. SUSD4a gene expression was analysed by qPCR using 10 ng cDNA/well as described above. Three different primers specific for SUSD4a (Hs01042141_m1, Hs01047294_m1, Hs01047293_m1), and the reference primers (Applied Biosystems) cyclophilin A, beta-2 microglobulin (B2M, Hs00984230_m1), and HPRT-1 were used.

The fresh or stimulated T cells were lysed in RIPA buffer and 10 μg total protein/sample was analysed by 12 % SDS-PAGE under reducing conditions. The gel was blotted and the membrane was stained as described above. After development of the SUSD4a signal, the lower part of the blot was washed and stained for the loading control B2M (17 ng/ml; Abcam, ab75853).